2021 3rd East Indonesia Conference on Computer and Information Technology (EIConCIT)
April 09-11. 2021, ISTTS Surabaya, Indonesia
978-1-6654-0514-0/21/$31.00 ©2021 IEEE 61
Application of Deep Learning for Early Detection of
COVID-19 Using CT-Scan Images
1
st
Judith Chrisolita Sangidong
Dept. of Information Technology
Universitas Kristen Satya Wacana
Salatiga, Indonesia
jchrisolita@gmail.com
2
nd
Hindriyanto Dwi Purnomo
Dept. of Information Technology
Universitas Kristen Satya Wacana
Salatiga, Indonesia
hindriyanto.purnomo@uksw.edu
3
rd
Fian Yulio Santoso
Dept. of Information Technology
Universitas Kristen Satya Wacana
Salatiga, Indonesia
fianyuliosantoso@gmail.coml
Abstract—COVID-19 pandemic caused a vast impact
worldwide. The imbalance between the number of tools for
COVID-19 detection and the demand for COVID-19 tests from
citizens has overwhelmed the government. To overcome this
problem, artificial intelligence is utilized, specifically in the deep
learning field. In this paper, we propose FJCovNet, a new deep
learning model based on DenseNet121. FJCovNet managed to
get an accuracy of 98.14%, surpassing Xception with an
accuracy of 84,24%, VGG19 with an accuracy of 95.25%, and
ResNet50 with accuracy of 91.53%. FJCovNet also managed to
get less training time with 612 seconds, lesser than VGG19 with
808 seconds and ResNet50 with 809 seconds, and only slightly
more than Xception with 609 seconds.
Keywords—COVID-19, CT-Scan, deep learning, pretrained
model.
I. INTRODUCTION
Coronavirus Disease (COVID-19) is an infectious disease
caused by the new coronavirus (Sars-CoV-2) which was found
in the city of Wuhan, Hubei Province, China, in 2019 and
spread throughout the world in 2020, resulting in COVID-19
being declared as a Pandemic by WHO on March 11, 2020 out
[1]. The unavailability of a vaccine for COVID-19 causes the
number of COVID-19 cases to continue to grow. As of March
3
rd
2021 10.40 WIB, cases of COVID-19 worldwide reached
21,623,570 positive patients, 91,115,487 patients recovered,
and 2,560,602 patients died [2].
The impact of COVID-19 is life-threatening and results in
a decline in all aspects of life. In the field of education,
teaching and learning activities are carried out online.
Students in remote areas experience difficulties due to limited
information technology infrastructure [3]. In the health sector,
there is a high demand for medical personnel and/or doctors
capable of dealing with COVID-19. Lack of PPE (Personal
Protective Equipment) supplies, heavy workloads, and stress
due to not being able to meet family and anxiety about
contracting COVID-19 have increased the number of medical
personnel and/or exhausted doctors who contracted COVID-
19 and even died [3], [4]. On the economic front, COVID-19
has caused most countries to face a recession in 2020 [5]. If
prevention is not carried out, the country can experience
poverty, which has a major impact on its welfare.
To maintain the welfare of the state and society, the
government raises funds to help referral hospitals for COVID-
19 and monitor individuals who have the potential to have
COVID-19 to minimize the spread of the virus. Therefore,
large funds are needed. The country's economy must keep
running. To ensure that the economy continues without
increasing the number of COVID-19 cases, the New Normal
concept is implemented to carry out normal activities by
implementing health protocols to prevent transmission of
COVID-19 [6].
An important step needed to minimize the spread of the
virus and implement the New Normal is a test to detect
whether someone has COVID-19. If an individual is positive
for COVID-19, then that individual must be isolated and
undergo treatment and not do activities. The test widely used
to confirm COVID-19 and trusted to be accurate is the
reverse-transcription polymerase chain reaction (RT-PCR).
However, there are times when the sensitivity of RT-PCR is
not very high for early detection and treatment in suspected
patients [7], [8].
This can increase the risk that individuals who are positive
for COVID-19 can continue their activities and endanger the
people they meet. Besides, the diagnosis of a suspect patient
requires several time-consuming tests [8]. The number of
suspect patients is always increasing, and the number of test
kits in hospitals is relatively limited. [9]. These things can
hinder the process of preventing COVID-19.
II. LITERATURE STUDY
Artificial Intelligence (AI) is studying how to program
machines or computers to do things humans do [10]. Along
with the development of the times, artificial intelligence
technology is increasingly sophisticated and is increasingly
being used because of its ability to do human work to increase
efficiency in everyday work. Artificial intelligence can be
developed into various types of fields of science. One of the
applications of artificial intelligence that has been developed
is in the field of Computer Vision. Computer Vision is a
branch of Artificial Intelligence that studies and empowers
machines to function like the human eye [11].
Computer Vision has been applied to various fields of
science. One of the studies in agriculture uses computer vision
to determine the quality of tomatoes [11]. Meanwhile, in the
health sector, Computer Vision is widely applied. Research
conducted by P Vávra et al. demonstrated an increasing
interest in surgeons to use AR in surgery as it increases the
safety and effectiveness of surgical procedures [12]. In a study
conducted by Shivangi Jain et al., Image processing was
applied to detect melanoma skin cancer [12], [13]. Computer
vision is also applied in the early detection of breast cancer
[14] and early detection of pneumonia [15]. The application
of computer vision in these studies is made with deep learning,
which is a type of artificial intelligence that uses an algorithm
or set of mathematical instructions inspired by how the human
brain works [16]. Deep learning has shown excellent
performance in solving computer vision challenges, such as
2021 3rd East Indonesia Conference on Computer and Information Technology (EIConCIT) | 978-1-6654-0514-0/20/$31.00 ©2021 IEEE | DOI: 10.1109/EIConCIT50028.2021.9431887